โ† All examples

ModelEthio-ASR Amharic (600M)2Ethio-ASR Multi (300M)9Ethio-ASR Multi (600M)4Ethio-ASR Multi (94M)8Meta MMS-1B8OmniASR-CTC (1B)2OmniASR-CTC (300M)3OmniASR-CTC (3B)2OmniASR-LLM (1B)7OmniASR-LLM (300M)3OmniASR-LLM (3B)1OpenAI Whisper-small8
Audio qualityClean Studio2Ambient Room Noise2Harsh Environment2Distant / Low-Volume Mic2Wideband Phone (VoLTE)6Randomized Phone Channel9Narrowband 2G + Packet Loss16

What did the model get wrong?

Ethio-ASR Multi (300M) Randomized Phone Channel   Example 15 of 20  ยท  speaker: Female

๐ŸŽง Listen โ€” Randomized Phone Channel
Phone-call audio with random extra distortions layered on top.
โœ… What was actually said
แŠฅแˆฝ แŠ แˆแŠ• แŠฅแŠ•แ‹ฐแ‰ต แŠแ‹ แ‹จแˆแŒ แ‰€แˆ˜แ‹ แŠ แ‹ตแˆต แ‰แŒฅแˆญ แŠ แˆแ‹ฐแˆจแˆฐแŠแˆ
๐Ÿค– What this model heard
tir แŠฅแˆฝ แŠ แ‰ฅ แˆแ‹ต แŠจ แˆ˜แŒ แ‰€แˆ แŠ แ‰ข แ‰แŒฅแˆซ แ‹ณ แ‰ณ
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
1
Correct
7
Wrong
0
Missed
2
Extra
12%
Words right
Out of 8 spoken words, this model got 1 right (12%). Errors: 7 words wrong; 2 extra.

Try switching Audio quality โ€” see the same model degrade as the audio gets noisier.